Autonomic resource management through self-organising agent communities

In this paper we describe a model of a decentralised system in which there may be no global information repository and the demand for particular resources changes over time. Agents must therefore organise the information that is available to them locally in order to adaptively respond to changes in demand in an efficient manner. Where previous studies have explored the role of system heterogeneity brought about by limiting knowledge [2] or using decision procedures that diversify agent behaviour [1], here we focus specifically on how varying the amount of information available to agents affects the flow of information. By conducting a thorough evaluation of our model, the results demonstrated that when agents possessed only a limited awareness of their peers within their neighborhood, they exhibited self-organising behaviour resulting in the emergence of community structures that support locally shared information. By limiting the quantity of information shared by constraining an agent's memory size, we found that a stable local community behavior emerged that was robust and efficient, and also adaptive when the whole agent population was exposed to global fluctuations in service demand.